Simulating a population split

By default vcfsim simulates a single panmictic population. To produce a VCF that contains two populations with a shared history, switch to two-population mode:

vcfsim \
    --chromosome 1 --replicates 1 --seed 1234 \
    --sequence_length 10000 --ploidy 2 --Ne 100000 --mu 1e-6 \
    --percent_missing_sites 0 --percent_missing_genotypes 0 \
    --sample_size 10 \
    --population_mode 2 --div_time 1000 \
    --output_file myvcf

The model

--population_mode 2 simulates a clean two-population split: an ancestral population C of effective size --Ne splits into two present-day populations A and B, both of size --Ne, at --div_time generations before present.

• --population_mode 2 activates the split.

• --div_time sets the split time in generations before present. Required when --population_mode 2; ignored otherwise.

Sample distribution

Samples are split evenly between the two populations, so the total sample count must be even. This applies to all three sample-specification flags:

• --sample_size 10 puts 5 samples in A and 5 in B.

• --samples A1 A2 A3 A4 puts the first two in A and the last two in B.

• --samples_file names.txt likewise splits the names in order.

If an odd number of samples is requested in mode 2, vcfsim raises an error rather than silently truncating.

Using the output for F_ST simulations

Two-population mode is the natural choice for benchmarking between-population statistics (F_ST, d_xy). The split time gives you a single, intuitive knob for the level of differentiation: larger --div_time produces deeper splits and higher F_ST.

When using the output for pixy-style analyses, build the populations file from the sample names emitted by vcfsim — the first half belong to population A and the second half to population B.